Friday, November 4, 2011

2011-2012 Season Preview: The Uber-Post

Princeton or Harvard. Harvard or Princeton.

Last year, that was pretty much the only question that mattered. This year, who will win the Ivy title is pretty much the only question that doesn't matter.

Of far greater importance are the debates about Yale as the number one contender, the fall of Princeton, the rise of Penn and the high hopes of Brown, Cornell and Columbia. With most pundits worrying far more about where Harvard will be seeded rather than if the Crimson will win the title, the goal of this year's projections will be to parse out the good Ivy teams from the decent and the decent from the bad.
The methodology employed to perform this task is the same as always. Build up from the individual level to the team to project each squad's offensive and defensive rating. Use the team strategy to guess at where the variance figures will fall. Then, simulate the entire season a bunch of times to get a range of possible outcomes. So, yes, there will be numbers to chew on. But I will cut them up into bite size pieces, don't worry.

Before we get started, while the league record projections are based upon Ivy team strengths forecasted in the model, the non-league record projections have Pomeroy's Ratings inserted for all opponents (like Colonel Jessup, I have neither the time nor the inclination to project all 300+ Division I teams). The assumption that goes into the non-conference record is that all of Pomeroy's misses are likely to be high or low in such a manner that those deviations mostly cancel out.

There are no surprises at the top, as the Crimson is the clear favorite to take home its first solo Ivy title and second NCAA bid in its history. Taking these numbers out to an extra decimal, the separation of second-place Yale (8.7 wins) from fifth-place Cornell (6.8 wins) is only 1.9 wins, meaning that there could be a lot of random movement within this group (more on probabilities later in this post).

If the results hold, the Ivies could see their third-straight year of three postseason invites (the Bulldogs declined their CBI invitation last year) though it would take some out-performance of expectations for either Yale or Princeton to build the profile necessary for NIT qualification. If Penn can find a way to land at 16-15 or 17-14, the Quakers could sneak into the CIT or CBI, if they'd be willing to pony up a hosting fee, as The Palestra would be a highly sought after venue by those tournaments.

Brown and Columbia landed a step behind the two through five slots, primarily due to weak projected defense. Offensively, the Bears and Lions should be able to keep up with all but Harvard, however, the inability to get stops on the other end will likely be what relegates them to the lower division in the end.

Finally, for those of you who are adept at mental math, the rounding on the Ivy wins causes a buildup to a record of 57-55. When I round off the decimals, this will often happen. Have no fear. When we dive deeper into the numbers later, you'll get a better sense of which teams are rounding up and down.

The above chart is the expected finish for each team with ties. Thus, your higher places will sum to more than 100 percent (more likely to be ties for 1st, 2nd, 3rd) and your lower places will sum to less than 100 percent (can't tie for 8th). The far right "Alloc1st" column merely allocates titles from potential playoffs to the involved teams on a pro rata basis. It can be thought of as the true odds of getting the Ivy League's NCAA bid.

Every team but Dartmouth finds its way to a share of the title in at least one of the simulations. Cornell is expected to lead the league in variance, which is why it covers so much ground in terms of potential finishes.

Despite Penn and Princeton being knotted at an 8-6 expectation, the Tigers are much closer to the Bulldogs' heels than the Quakers are to them. Princeton and Yale take almost 80 percent of the bids when the Crimson fails to do so.

And that brings us to Harvard. At 90.6 percent to defend their title and 85.1 percent to claim their second NCAA bid, they're an overwhelming favorite, but whether one should use the term "lock" is debatable. At only 80.3 percent to win the league outright, that leaves one-in-five times where the Crimson is either on the outside looking in or back in another playoff scenario.

During the mid-2000s, when most of the league was treading water outside the Top 200, a dominant team could realistically plan on cutting the nets in February. Last year, a majority of the league (5/8) landed in the Top 200 and that's likely to be repeated or exceeded this season. With fewer and fewer gimmes on the schedule, the road for the favorite becomes tougher and tougher.

The above chart lays out the probability of each team reaching a certain number of Ivy wins. The most important points are that Harvard is relatively unlikely to go undefeated in the league slate and that Dartmouth is very unlikely to go winless.

Another important point is that the spreads in projected win totals are very wide, owing primarily to the parity among the middle six teams. There will be a ton of Ivy games in the 30-50 percent win probability range, which could lead to some streaks of luck that send teams far off their expectations. That's why a team like Cornell can finish with four or fewer wins nine percent of the time while still having a 15 percent chance to hit at least nine wins under the same average expectation.

It should be a wild year in the middle of the pack, and if one of those teams catches all the breaks in their close games, it could become a wild year at the top as well.

The Ivy League has improved by leaps and bounds offensively over the past few years. Each team (aside from Dartmouth, of course) has at least a couple of weapons capable of scoring the ball efficiently. Throwing out the Big Green and frontrunner Harvard, all of the other six teams are within five points per possession offensively.

The separating factor in the Ivies this year will be defensive stops. Brown, Columbia and Cornell struggled to get them last year, and it's hard to see how they've improved on that side of the ball. Each lost their best defensive rebounder and have plenty to prove when it comes to interior defense.

Princeton should continue its regression to the mean in the absence of Kareem Maddox, while Penn will likely improve just by its opponents not shooting 76 percent from the stripe. Whether the Quakers fulfill the rest of their defensive promise depends on how their injury situations shake out in the frontcourt.

With Greg Mangano and Jeremiah Kreisberg anchoring the paint, the Bulldogs should be in the discussion with the Crimson for the league's best defense. Yale does an amazing job of forcing misses and turning those into stops. It doesn't generate a ton of turnovers, though, which is really the only thing holding it back.

The final note on the Pomeroy Ratings is that non-conference really matters. How the Ivy League performs in November and December determines the size of the pie for which the member teams will be playing. Thus, if every Ivy team has a laundry list of injuries akin to those announced by Cornell and Penn, the league-wide rating might take a beating. Since the goal of these projections is to get Ivy play right, even the injuries that are already known have not been factored into the ratings, unless they are believed to carry into the Ivy slate. So, while Penn might play like a 155 in Ivy play, if much of its regular rotation misses games in November and December, its ultimate ranking might come up far short of that.

Last but not least, it's the game-by-game projections. These will be updated on the Projections Page of this blog throughout the season (as teams and opponents rise and fall off of their preseason projections).

The scores are derived by multiplying the expected Offensive and Defensive Ratings by the opponent's rating to get the per-possession scoring for each team. Then, a pace estimation is made to arrive at a final score. The percentages in parentheses are always the home team's odds of winning (in the case of a neutral site game, you'll be able to tell who the percentage applies to based on who is predicted to win the contest). Those percentages are the result of a Bayesian formula incorporating both average expectations and variance.

Keep in mind that a team's predicted record is not merely the sum of all the games it is expected to win and those which it is expected to lose. A simple illustration is to take two scenarios, one in which a team has a 99.99 percent chance of winning in four different games of the time and another in which those odds are consistently at 50.01 percent. The teams in each example will be favorites in all four games, but the expected record for the former would be 4-0, while the expected record for the latter would be 2-2.

Thus, despite Harvard being a favorite in each of its Ivy games, it's expected cumulative record is 12-2, because it faces four games where its odds of winning are around 70 percent or less and a few more where its odds of winning fail to reach 90 percent.


The most intriguing part of this year's campaign will be how the middle of the league shakes out.

Will the optimistic Penn faithful, which saw their team vault into the Top 200 for the first time since last winning the Ivy title in 2007, be rewarded with another forward step, or will the loss of Jack Eggleston leave them stuck running in place, or worse, tumbling back down the ladder? (Likely a small step forward, but sideways is possible).

Can Princeton hold it together after having lost Dan Mavraides, Kareem Maddox and Coach Sydney Johnson to claim the spot of number one contender, or will the Tigers enter a minor rebuilding phase? (Look out below - it could be a much larger fall than just 20 or 30 Pomeroy spots).

Is Yale ready to lead the chase pack behind a stellar frontcourt and decent guard play? (My Sources Say Yes.)

Can any of Brown, Columbia and Cornell bust through what seems to be a decently-consensus top four to finish in the league's upper division? (Hot shooting on the right nights from the Big Red could be the difference).

That Harvard is the team to beat is hardly news, nor is it interesting, though the fact that it is only 80 percent to win the league outright might raise some eyebrows especially among those who have already Sharpied the Crimson into a 40-49 spot on the S-Curve. The intriguing part will be finding out which teams have the young talent to take a stronger shot at a Harvard team that will lose two starters (including the 2011 Ivy POY) heading into the 2012-2013 season.


  1. So far, one correction: Temple plays at Penn, not the other way around as I have it. That would make that game Temple 70, PENN 62 (21%).